- Email: [email protected]
A Pessimistic Explanatory Style Is Prognostic for Poor Lung Cancer Survival
ORIGINAL ARTICLE
A Pessimistic Explanatory Style Is Prognostic for Poor Lung
Cancer Survival Paul Novotny, MS,* Robert C. Colligan, PhD,† Daniel W. Szydlo, BA,* Matthew M. Clark, PhD,†
Sarah Rausch, PhD,† Jason Wampfler, BS,‡ Jeff A. Sloan, PhD,* and Ping Yang, MD, PhD§
Background: Several studies have demonstrated the importance of personality constructs on health behaviors and health status. Having a pessimistic outlook has been related to negative health behaviors and higher mortality. However, the construct has not been well explored in cancer populations. Methods: Survival time of 534 adults who were diagnosed with lung cancer was examined. The patients had completed the Minnesota Multiphasic Personality Inventory approximately 18.2 years before receiving their lung cancer diagnosis. Minnesota Multiphasic Personality Inventory Optimism-Pessimism scores were divided into high (60 or more) and low scores (�60), and log-rank tests and Kaplan-Meier curves were used to determine survival differences. Multivariate Cox models were used for assessing prognostic values of pessimism along with other known predictors for lung cancer survival outcome. Bootstrapping of the survival models was used as a sensitivity analysis. Results: At the time of lung cancer diagnosis, patients were at an average age of 67 years old; 48% of them were women, 85% had non-small cell lung cancer, 15% had small cell lung cancer, 30% were stage I, 4% were stage II, 31% were stage III/limited, and 35% were stage IV/extensive. Patients who exhibited a nonpessimistic explanatory style survived approximately 6 months longer than patients classified as having a pessimistic explanatory style. Conclusion: Among lung cancer patients, those having a pessimistic explanatory style experienced a less favorable survival outcome, which may be related to cancer treatment decisions. Further research in this area is warranted. Key Words: Explanatory style, Optimism, Pessimism, Lung cancer, MMPI, Survival.
M
“
ind-body” relationships have been revered since the time of Socrates. Personality or emotional factors may have a direct impact on physiological states or mind, which are evident in the current emphasis on stress reduction, relaxation, meditation, and activities related to disease pre vention and wellness promotion. Indeed, psychosocial factors may be predictive of poor disease outcome, including cancer survivorship. Pessimism and optimism are personality con structs that have been shown to be important in the general population and in some medical populations.1,2 Having a pessimistic explanatory style means the individual attributes bad events to internal, stable, and global causes; he or she tends toward self-blame, fatalism, and catastrophic thinking.3 Recently, the role of having a pessimistic explanatory style has been explored in several cancer populations. For exam ple, Kung et al.4 found that having a pessimistic explanatory style was associated with poor quality of life in head and neck and thyroid cancer survivors; Petersen et al.5 found that a pessimistic explanatory style was predictive of poor quality of life in 268 breast cancer survivors. In contrast, optimism has also been associated with positive health outcomes in a number of studies; for instance, higher levels of optimism have been associated with lower blood pressure during daily life,6 better recovery from coronary artery bypass surgery,7 and longer survival in patients with head and neck cancer.8 However, such factors may be of little importance when the body experiences serious or fatal medical conditions, or potentially lethal diseases such as lung cancer. Using Selig man’s theory of causal attribution,3 the current retrospective study examined the relationship between a pessimistic ex planatory style and survival in a group of patients diagnosed with lung cancer.
(J Thorac Oncol. 2010;5: 326–332)
PATIENTS AND METHODS Study Design and Patients *Cancer Center Statistics, †Department of Psychiatry and Psychology, ‡Di vision of Biomedical Statistics and Informatics, and §Division of Epide miology, Mayo Clinic College of Medicine, Rochester, Minnesota. Disclosure: The authors declare no conflict of interest. Address for correspondence: Ping Yang, MD, PhD, Division of Epidemiol ogy, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail: [email protected] Copyright © 2010 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/10/0503-0326
326
This study is a retrospective, observational cohort design aimed at examining the relationship between ex planatory style, as measured by scores on the OptimismPessimism (PSM) scale of the Minnesota Multiphasic Personality Inventory (MMPI), and survival in patients diagnosed with lung cancer. Participants in this study were patients with primary lung cancer who were enrolled into the Epidemiology and Genetics of Lung Cancer Research Program at Mayo Clinic Rochester since 1997, in which all patients at Mayo Clinic who were diagnosed with lung
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
cancer have been offered enrollment in a prospective cohort study.9 –12 All patients enrolled provided informed consent, and the study has been approved by the Institu tional Review Board. Trained study personnel reviewed the medical records; all patients completed health-related surveys when they entered the cohort study and again at 6 and 12 months and were mailed similar surveys on an annual basis. Information on demographics, previous or concurrent illnesses, tobacco usage and exposure, tumor staging, nutritional habits, and cancer therapy were ab stracted and entered into the database. Comorbidities were combined into three variables for having any other lung disease, any other cancer, and any other disease. Having another lung disease included asthma, chronic bronchitis, emphysema, chronic obstructive pulmonary disease, pneu monia, tuberculosis, or cystic fibrosis. Any other cancer was defined as a cancer diagnosis except lung cancer. Any other disease was defined as any disease except cancer or a lung disease; this category included diabetes, heart disease, arthritis, anemia, migraines, drug addiction, and any other medical problem. For a more detailed discussion of variable definitions, see Ref. 9. The Revised Staging System of non-small cell lung cancer (NSCLC) was used.14 All cancer treatment decisions were deferred to the individual patient’s healthcare providers, and enrollment into the research program did not in any way influence clinical decision making. This study used retrospective data from 534 individuals diagnosed with lung cancer who had completed the MMPI before receiving their lung cancer diagnosis. Medical records included information on age, sex, disease stage, smoking status, and survival status. If patients completed more than one MMPI, the PSM score from the earliest MMPI completed before lung cancer diagnosis was used to assess PSM. The MMPI was com pleted at the request of the patient’s physicians as part of the patient’s medical care or as a participant in a large research study. Patients who were asked to complete an MMPI were more likely among the most complex of referrals at our tertiary care medical center, because the physician making the request believed there were psycho logic factors intertwined with the presenting symptoms.
Measures Smoking Status Patients who reported smoking fewer than 100 ciga rettes in their lifetime were classified as “never smokers.” Patients who had not smoked a cigarette in the past 30 days were classified as “former smokers,” and patients who re ported smoking any cigarette in the past 30 days were classified as “current smokers.”15
The Minnesota Multiphasic Personality Inventory The original MMPI consisted of 550 unique true/ false items about thoughts, feelings, attitudes, physical symptoms, emotional symptoms, and previous life experi ences.16 The original MMPI and its current revision, the MMPI-2,17 have been the most widely used and thoroughly researched of the self-report measures of personality func
Pessimistic Explanatory Style for Lung Cancer
tioning.18 All the analyses in the article were based on the original MMPI. The PSM scale for the MMPI was derived from Seligman’s theory of explanatory style19 and constructed from the items of the MMPI item pool.20 According to Seligman’s theory, the manner in which individuals ex plain the cause of significant life events (both good and bad events) exerts considerable influence over three as pects of their future physical and mental health, which include decreased quality of physical health, increased risk for depression, and reduced occupational or aca demic achievement.21 More specifically, Seligman’s the ory postulates that people who (1) attribute the causes of adverse events in their lives to themselves (i.e., an interval explanation, “It’s me . . .”), (2) carry the expectation that the condition will persist (i.e., a stable explanation, “. . . happened again, as usual . . .”), and (3) that it will affect other aspects of their functioning (e.g., a global explanation, “. . . and now my life will be ruined; I’ll never get to . . .”) are at risk for undermining their subse quent physical health, emotional and mental functioning, and life achievement. The PSM scale yields normalized T-scores (mean � 50, SD � 10). Lower PSM scores indicate having an optimistic explanatory style; whereas, higher scores indi cate a pessimistic explanatory style. Similar to previous research, PSM scores were divided into two groups for analysis: PSM scores representing the nonpessimistic (e.g., optimistic) patients were in the first group (PSM scores of �60). Subjects with PSM scores of 60 or higher (1 SD above the mean) were considered to have a pessimistic explanatory style.4,5 Although this cut point has been used in other studies of pessimists,1,2,4 it is not based on a clinical cutoff score. We used recursive partitioning22 to determine the appropriateness of the score as our cut point when modeling survival times; the optimal cut point for our study cohort fell at 58, but we used a normalized score of 60 as being nearly optimal for the purpose of compa rability with published literature.
Analyses PSM scores were divided into pessimistic explana tory style or nonpessimistic explanatory style, and t tests were used to determine demographic differences between patients highest and lowest in PSM scores. Log-rank tests and Kaplan-Meier curves were used to determine survival differences. Stepwise Cox proportional hazards models were used to model survival adjusting for age, gender, smoking status, cancer type, stage, and comorbidities. Variables were entered into the model one at a time. Modeling stopped when no other variables met the 0.05 significance level for entry into the model. Backward modeling (starting with a saturated model and dropping one variable at a time) was used to confirm the results of the stepwise model. Martingale residuals were explored to determine the functional form of the PSM variable, and diagnostic checks were conducted to verify the fit of the final models. Bootstrapping of the survival analysis models was carried out as a sensitivity test of our findings by
Copyright © 2010 by the International Association for the Study of Lung Cancer
327
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Novotny et al.
TABLE 1. Selected Characteristics of Patients with Lung Cancer Who Completed the MMPI Before Being Diagnosed with Lung Cancer Characteristicsa Age at DXc Mean (SD) Stage I II III/limited IV/extensive Cancer type SCLCd NSCLCe Median time from diagnosis to last follow-up
MMPIb (N � 534)
No MMPIb (N � 8786)
67.4 (10.39)
65.7 (11.14)
147 (30) 21 (4) 152 (31) 176 (36)
1999 (23) 589 (7) 2659 (31) 3310 (39)
72 (15) 424 (85) 14.8 mo
933 (11) 7624 (89) 15.5 mo
a
Unless specified, number and percentage in parentheses are presented. MMPI, Minnesota Multiphasic Personality Inventory; c DX, Diagnosis; d SCLC, small cell lung cancer; eNSCLC, non-small cell lung cancer. b
generating 10,000 random samples with replacement from our observed dataset.23
RESULTS A total of 534 subjects had completed a MMPI before receiving their lung cancer diagnosis between 1997 and 2006. Table 1 shows patients with and without a MMPI. There is no indication that patients with MMPI scores were clinically different on these characteristics than patients who did not complete a MMPI, and explan atory style is unlikely to have biased completion rates for the MMPIs among medical outpatients.24 Table 2 provides baseline demographics for the 534 patients who form the basis of this report. On average, patients had completed the MMPI 18.2 years before being diagnosed with lung cancer. Most patients had NSCLC (85%), and 34% of patients with NSCLC had early stage (I or II) lung cancer. In addition, most patients were current or former smokers (80%). None of the demographic variables were significantly different between patients with a pessimistic explanatory style and patients with a nonpessimistic explanatory style. There were 110 patients who completed more than one MMPI before their lung cancer diagnosis. Of the 110 patients with multiple MMPI scores, 12 (11%) went from being classified as pessimists to nonpessimists and 13 (12%) changed from nonpessimists to pessimists. The median change for the patients was a decrease of one-half point on their PSM scores. We used the first PSM score reported before lung cancer diagnosis to achieve a greater temporal span between the PSM score and cancer diagno sis. This minimizes the likelihood that a PSM score was affected by proximity to the personal stresses of receiving a diagnosis of lung cancer and the significant symptoms of illness preceding the cancer diagnosis related to the cancer.
328
TABLE 2. Characteristics of 534 Patients with Lung Cancer Who Completed a MMPI
Characteristicsa Age at MMPIb Mean (SD) Median Range Age at DXc Mean (SD) Median Range Years From MMPIb to DX Mean (SD) Median Range Smoking status Missing Never smoker Former smoker Current smoker Current or former smoker Gender Female Male Cancer type SCLCe NSCLCf Stage Missing I II III/limited IV/extensive Any surgery Yes No Any chemotherapy Yes No Any radiation therapy Yes No Any other cancer Yes No Any other lung disease Yes No
Nonpessimistic Pessimistic Explanatory Style Explanatory PSMd <60 Style PSMd >60 (n � 304) (n � 230)
Total (n � 534)
48.4 (12.94) 49 14–97
49.9 (12.49) 51 15–78
49.0 (12.76) 50.0 14–97
67.6 (11.10) 68.5 34–98
67.0 (9.40) 68.0 39–88
67.4 (10.39) 68 34–98
19.0 (10.08) 18.6 0.01–43.5
17.1 (9.38) 16.6 0.02–43.4
18.2 (9.82) 17.5 0.01–43.5
1 48 (16) 117 (39) 117 (39) 21 (7)
2 21 (9) 89 (39) 103 (45) 15 (7)
3 69 (13) 206 (39) 220 (41) 36 (7)
149 (49) 155 (51)
108 (47) 122 (53)
257 (48) 277 (52)
37 (13) 243 (87)
35 (16) 181 (84)
72 (15) 424 (85)
24 82 (28) 10 (4) 94 (34) 94 (34)
14 65 (30) 11 (5) 58 (27) 82 (38)
38 147 (30) 21 (4) 152 (31) 176 (36)
123 (40) 181 (60)
75 (33) 155 (67)
198 (37) 336 (63)
85 (28) 219 (72)
62 (27) 168 (73)
147 (28) 387 (72)
69 (23) 235 (77)
40 (17) 190 (83)
109 (20) 425 (80)
32 (11) 272 (89)
34 (15) 196 (85)
66 (12) 468 (88)
61 (20) 243 (80)
53 (23) 177 (77)
114 (21) 420 (79)
a
Unless specified, number and percentage in parentheses are presented. MMPI, Minnesota Multiphasic Personality Inventory; c DX, Diagnosis; PSM, Optimism-Pessimism Scale Score; eSCLC, small cell lung cancer; f NSCLC, non-small cell lung cancer. b
d
Copyright © 2010 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
TABLE 3.
Median Survival (mo) by PSM Groups 95% Confidence Interval
Overall population PSMa �60 (nonpessimistic) 60 � PSMa (pessimistic) a
Median
Lower
Upper
14.8 19.2
13.4 14.8
18.7 23.5
13.1
10.9
14.8
TABLE 4.
5-Year Log-Rank Survival p Rate 0.0105
27.8% 32.9% 21.1%
PSM, Optimism-Pessimism Scale Score.
Pessimistic Explanatory Style for Lung Cancer
Univariate Cox Models for Survival
Variable Pessimist (60�) Age at diagnosis Male Stage II Stage III/limited Stage IV/extensive Former smoker Current smoker Current or former smoker Chemotherapy Radiation Surgery Other lung treatment Any cancer Any lung disease Any other disease
95% Hazard Ratio Confidence Limits
p
Hazard Ratio
0.0108 �0.0001 0.0141 0.1500 0.2087 �0.0001 0.0452 0.0045 0.0034
1.305 1.022 1.292 0.655 1.155 4.134 0.805 1.346 1.706
1.064 1.012 1.053 0.368 0.923 3.305 0.651 1.097 1.193
1.601 1.033 1.585 1.165 1.445 5.171 0.995 1.653 2.441
0.0058 0.0253 �0.0001 0.2965
1.365 1.317 0.234 1.601
1.094 1.035 0.182 0.662
1.703 1.676 0.300 3.870
0.0675 0.0012 �0.0001
0.748 0.655 0.391
0.549 0.506 0.293
1.021 0.846 0.521
TABLE 5. Multivariate Cox Model for Survival Including Pessimism, Treatment, and Comorbidities
Variable
FIGURE 1. Kaplan-Meier curve of survival time by PSM groups. Patients with a pessimistic explanatory style have shorter survival times than other patients.
Pessimistic Explanatory Style and Survival Table 3 shows median survival times (in months) and the 5-year survival rate for all 534 patients and compares patients with low versus high scores on the PSM scale. Patients (both women and men) in the nonpessimistic category survived about 6 months longer compared with patients with a pessimistic explanatory style, as shown in the Kaplan-Meier curve in Figure 1. Five-year survival rates for the two groups were 32.9% for nonpessimists and 21.1% for pessimists. The results from univariate Cox models are shown in Table 4. Stage of disease and treatment were highly correlated, i.e., 95% of patients with stage II cancer had surgery, 30% of patients with stage III had surgery, and only 6% of patients with stage IV had surgery. Because there was strong collinearity between stage and treatment, only treatment was used in the multivariate models along with other covariates. Results from the stepwise multivar iate model using pessimists versus nonpessimists are shown in Table 5, confirming that patients with a pessi mistic explanatory style have significantly worse survival rates even after adjusting for other known prognostic variables (hazard ratio � 1.25, p � 0.03). Having comor bidities seem to be associated with significantly better survival (Table 4); however, the effects are confounded by
Pessimist (60�) Age at diagnosis Male Any surgery Any radiation therapy Any chemotherapy Current smoker Some smoking history Any other cancer Any other lung disease
p
Hazard Ratio
0.0339 �0.0001 0.02 �0.0001 0.17 0.70 0.0005 0.0005 0.35 0.66
1.25 1.04 1.28 0.23 1.22 0.95 1.54 2.02 1.18 1.07
95% Hazard Ratio Confidence Limits 1.02 1.03 1.04 0.17 0.92 0.73 1.21 1.36 0.83 0.80
1.55 1.05 1.58 0.31 1.63 1.24 1.96 3.01 1.66 1.42
age, smoking status, and treatment. In particular, patients with comorbidities were more likely to have surgery and more likely to be current smokers. Patients with other cancers were on average about 5 years older than patients without other cancers. After adjusting for these confound ing effects as presented in Table 5, the effects of comor bidities are no longer significant. A stratified analysis was used to assess the relationship between pessimism and stage. Results suggest that having a pessimistic explanatory style is prognostic for survival in patients with stage I or II cancer but is not prognostic for patients with stage III or IV cancer. The observation was supported by a multivariate Cox model using only patients with stage I or II lung cancer (Table 6). After adjusting for demographics, comorbidities, and treatment, having a pes simistic explanatory style was still significantly related to
Copyright © 2010 by the International Association for the Study of Lung Cancer
329
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Novotny et al.
TABLE 6. Multivariate Cox Model for Survival for Patients with Stage I or II Cancer
Variable Pessimist (60�) Age at diagnosis Male Any surgery Any radiation therapy Any chemotherapy Current smoker Some smoking history Any other cancer Any other lung disease
95% Hazard Ratio Confidence Limits
p
Hazard Ratio
0.02 0.001 0.17 0.01 0.10 0.64 0.03 0.10
1.91 1.06 1.47 0.39 1.95 1.21 1.87 2.52
1.10 1.02 0.85 0.19 0.87 0.53 1.06 0.84
3.29 1.10 2.53 0.81 4.36 2.76 3.31 7.57
0.33 0.25
0.72 1.42
0.38 0.79
1.38 2.56
shorter survival times (log-rank test p � 0.02, hazard ratio � 1.91). Several different sensitivity analyses provided results that were similar to the original analyses. A simulation involving 10,000 bootstrapped samples showed the robust ness of the results. Seventy-two percent of the bootstrapped samples resulted in significant p values, with nonpessimists having longer survival times than pessimists. A small minor ity of samples (0.4% or 42 out of 10,000 samples) had results that showed pessimists having longer survival times, but the survival differences were not significant in all of those samples. Another sensitivity analysis used the last MMPI score reported before diagnosis instead of the first reported MMPI. In this analysis, the log-rank p value for differences in survival increased slightly and the difference in median survival decreased from 6.0 to 5.3 months. The relationship between depression (as measured by the MMPI, Scale 2, Depression) and survival was also ex plored, but there was no indication that depression was related to survival time (data not shown).
DISCUSSION This retrospective study examined the relationship be tween survival and explanatory style using scores from the PSM scale of the MMPI. The major finding in the study was that explanatory style obtained from a large sample of pa tients diagnosed with lung cancer and measured (on average) 18 years before receiving their lung cancer diagnosis, was statistically significant and clinically relevant for survival. More specifically, patients classified as having an optimistic or a nonpessimistic explanatory style survived an average of 6 months longer compared with the patients with a pessimis tic explanatory style. Furthermore, the relationship was inde pendent of smoking status, cancer stage, treatment, comor bidities, age, and gender. In examining stage of cancer, the prolonged survival time was only sustained among patients with stage I or II lung cancer, which indicates that if a patient has advanced disease,
330
then the potentially protective aspect of a nonpessimistic explanatory style can be overwhelmed by the severity of the disease process. However, the results support the premise that if a patient is diagnosed with lung cancer, and if the patient has a pessimistic explanatory style, the patient may be less likely to have surgery and may live 6 months less compared with patients with a nonpessimistic style. This may be due to the overall risk for comorbid medical problems (i.e., as predicted by Seligman’s theory of pessimistic causal attribu tion), which in turn decreases the likelihood of receiving a surgical intervention. However, if a patient’s lung cancer was identified in the early stage, and if the patient has a nonpes simistic explanatory style, then the patient’s estimated sur vival is significantly improved, in part, because surgical resection of the tumor can be completed as a curative treat ment. This 6-month potential benefit related to an optimistic explanatory style is more impressive when one considers that the median survival time for this patient population with lung cancer is �1 year.9 Seligman’s theory of explanatory style has been shown to be predictive of health status, mortality, health behaviors, and quality of life among general medical outpatients. The examination of a possible relationship between a pessimistic explanatory style and survivorship in oncology populations is a relatively new and provocative area of investigation. Such studies have yielded mixed results. Some suggest that having a pessimistic explanatory style before receiving a cancer diagnosis might be predictive of survival time and immune function; whereas, others have not found such an association. Rausch et al.25 found that higher levels of optimism were associated with improved immune function in women newly diagnosed with early stage breast cancer. Even in healthy subjects, researchers have found optimism to be associated with higher immune parameters, including higher T-lympho cyte numbers and natural killer cell activity.26,27 In a popu lation of black women coinfected with human immunodefi ciency virus and human papillomavirus, Byrnes et al.28 found that women who were more optimistic had better immunity. Specifically, greater optimism was related to higher natural killer cell cytotoxicity and cytotoxic/suppressor cell numbers after controlling for presence/absence of human papillomavirus, behavioral/lifestyle factors, and subjective impact of negative life events. In a recent review article, Coyne et al.29 noted mixed results from studies (often flawed by small samples and numerous potential confounders) examining broad person ality factors related to disease progression and mortality among patients with cancer. In summarizing the literature, Coyne et al. decry the lack of substantial support for widely held beliefs about the impact of personality factors, using a research model of their own large-scale work in head and neck cancers. Contrary to the view of Coyne et al., we deliberately chose not to use a broad measure of emotional well being. Rather, we used a theory-derived measure characterizing an enduring psychologic and cog nitive construct known as causal attribution. A summary of our approach follows:
Copyright © 2010 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Originally, Mayo Clinic researchers collaborated with Seligman and, using Seligman’s procedures for analyzing expository text, developed and published a bidirectional scale of PSM derived from the item pool of the original MMPI. Thus, Seligman’s theory of causal attribution was operation alized in the constructs of durable “pessimistic” and “opti mistic” personality traits. According to Seligman’s theory, individuals with a pessimistic explanatory style are at signif icant risk for later problems in three important areas of life functioning: (a) greater likelihood of adverse medical condi tions, (b) proneness toward mental health issues (particularly depression), and (c) reduced achievement (either occupa tional or academic). Seligman’s theory and the PSM scale have been vali dated among general medical outpatients. For example, a pessimistic explanatory style was significantly associated with increased mortality among medical outpatients who completed the MMPI approximately 30 years before follow up.1 Overall, results from several previous studies30 –34 have added considerable support to Seligman’s theory and the potentially adverse impact of having a pessimistic explana tory style. In general, we have found, as the theory predicts, patients classified as having a pessimistic explanatory style are at risk for poorer medical outcomes; whereas, being in the nonpessimist range of scores seems to be a psychologic protective factor. Seligman’s theory predicts that patients with lung cancer who were classified as having a pessimistic explanatory style would be at risk for gradually accruing adverse medical conditions over time, which may be related to the decreased likelihood of surgical treatment for patients in this explanatory category. Note: we are not implying a causal relationship between a pessimistic explanatory style and the development of lung cancer. Some limitations of the study should be considered when interpreting the results. First, the retrospective design of this study does not ensure that the measures were given at consistent times or to all possible participants. Patients who completed a MMPI might have exhibited behaviors or coping styles different from individuals who did not complete the MMPI; thereby, introducing selection bias into the study sample. Another potential source of bias relates to the mea surement of PSM at only one point in time. There is uncer tainty about whether a relationship existed between explan atory style and survival before diagnosis. A related limitation is that information was not available on cancer recurrence or comorbid medical conditions that may have occurred since the time of cancer diagnosis and could impact survival. Therefore, longitudinal studies are recommended to further investigate these possibilities. A final limitation is the gener alizability of the results; the sample was primarily white and only consisted of patients diagnosed with lung cancer, there fore, the results may not apply to more diverse populations or other types of cancer. Despite the limitations, our study adds to the grow ing literature on explanatory style (e.g., PSM) and survival in the general population, in medical populations, and specifically in patients with lung cancer. Future investiga tions may benefit from designing and testing interventions,
Pessimistic Explanatory Style for Lung Cancer
which address enhancing positive aspects of explanatory style and evaluating the potential physiological mecha nisms responsible for increased survival. For example, patients diagnosed with cancer could learn cognitive be havioral techniques to challenge negative thinking patterns and engage in effective, accurate problem solving. This may ultimately aid in enhancing current approaches to patient care, such that clinicians may improve survival not only by developing new medical treatments but also by targeting patients’ psychosocial characteristics most likely to negatively affect cancer treatment decisions and ulti mate outcomes.
ACKNOWLEDGMENTS This work was supported by U.S. National Institutes of Health grants, R01 CA 115879 and R01 CA, 84354, awarded to Ping Yang, MD, PhD. The authors thank Susan Ernst, MA, for her technical assistance with the article. REFERENCES 1. Maruta T, Colligan RC, Malinchoc M, et al. Optimists vs pessimists: survival rate among medical patients over a 30-year period. Mayo Clin Proc 2000;75:140 –143. 2. Maruta T, Colligan RC, Malinchoc M, et al. Optimism-pessimism assessed in the 1960s and self-reported health status 30 years later. Mayo Clin Proc 2002;77:748 –753. 3. Peterson C, Seligman ME. Causal explanations as a risk factor for depression: theory and evidence. Psychol Rev 1984;91:347–374. 4. Kung S, Rummans TA, Colligan RC, et al. Association of optimismpessimism with quality of life in patients with head and neck and thyroid cancers. Mayo Clin Proc 2006;81:1545–1552. 5. Petersen LR, Clark MM, Novotny P, et al. Relationship of optimismpessimism and health-related quality of life in breast cancer survivors. J Psychosoc Oncol 2008;26:15–32. 6. Raikkonen K, Matthews KA, Flory JD, et al. Effects of optimism, pessimism, and trait anxiety on ambulatory blood pressure and mood during everyday life. J Pers Soc Psychol 1999;76:104 –113. 7. Scheier MF, Matthews KA, Owens JF, et al. Optimism and rehospital ization after coronary artery bypass graft surgery. Arch Intern Med 1999;159:829 – 835. 8. Allison PJ, Guichard C, Fung K, et al. Dispositional optimism predicts survival status 1 year after diagnosis in head and neck cancer patients. J Clin Oncol 2003;21:543–548. 9. Yang P, Allen MS, Aubry MC, et al. Clinical features of 5,628 primary lung cancer patients: experience at Mayo Clinic from 1997–2003. Chest 2005;128:452– 462. 10. Sun Z, Aubry MC, Deschamps C, et al. Histologic grade is an indepen dent prognostic factor for survival in non-small cell lung cancer: an analysis of 5018 hospital- and 712 population-based cases. J Thorac Cardiovasc Surg 2006;131:1014 –1020. 11. Sugimura H, Yang P. Long-term survivorship in lung cancer. A review. Chest 2006;29:1088 –1097. 12. Jatoi A, Williams B, Nichols FC, et al. Is voluntary vitamin and mineral supplementation associated with better outcome in non-small cell lung cancer? Results from the Mayo Clinic Lung Cancer Cohort. Lung Cancer 2005;49:77– 84. 13. Visbal AL, Williams BA, Nichols FC, et al. Gender differences in non-small cell lung cancer survival: an analysis of 4,618 patients diagnosed between 1997–2002. Ann Thorac Surg 2004;78:209 –215. 14. Mountain CF. Revisions in the international system for staging lung cancer. Chest 1997;111:1710 –1717. 15. Clark MM, Croghan IT, Reading S, et al. The relationship of body image dissatisfaction to cigarette smoking in college students. Body Image 2005;2:263–270.
Copyright © 2010 by the International Association for the Study of Lung Cancer
331
Novotny et al.
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
16. Hathaway S, McKinley J. The Minnesota Multiphasic Personality In ventory. Minneapolis, Minnesota: University of Minnesota, 1943. 17. Butcher JN, Dahlstrom WG, Graham JR, et al. Manual for Adminis tration and Scoring: MMPI-2, Minnesota Multiphasic Personality Inventory-2 (Revised Edition). Minneapolis: University of Minne sota, 2001. 18. Butcher JN. A Beginner’s Guide to the MMPI-2, 2nd Ed. Washington, DC: American Psychological Association, 2005. 19. Gillham JE, Shatte AJ, Reivich KJ, et al. Optimism, pessimism, and explanatory style. In EC Chang (Ed.), Optimism and Pessimism: Impli cations for Theory, Research, and Practice. Washington, DC: American Psychological Association, 2001. 20. Colligan RC, Offord KP, Malinchoc M, et al. CAVEing the MMPI for an optimism-pessimism scale: Seligman’s attributional model and the assessment of explanatory style. J Clin Psychol 1994;50: 71–95. 21. Seligman MEP. Explanatory style: predicting depression, achievement, and health. In MD Yapko (Ed.), Brief Therapy Approaches to Treating Anxiety and Depression. New York, NY: Brunner, 1989. 22. Therneau TM, Atkinson EJ. An Introduction to Recursive Partitioning Using the Rpart Routines. Technical Report #61. Rochester: Mayo Clinic, 1997. 23. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. London: Chapman & Hall, 2003. 24. McLeod TG, Costello BA, Colligan RC, et al. Personality characteristics of healthcare satisfaction survey non-respondents. Int J Health Care Qual Assur 2009;22:145–156. 25. Rausch SM, Auerbach SA, McCain NL, et al. The relationship between psychosocial and immune variables in American women with breast cancer. In the 10th International Congress of Behavioral Medicine Meeting. Tokoyo, Japan, 2008. 26. Cohen F, Kearney KA, Zegans LS, et al. Differential immune system
332
changes with acute and persistent stress for optimists vs pessimists. Brain Behav Immun 1999;13:155–174. 27. Segerstrom SC, Taylor SE, Kemeny ME, et al. Optimism is associated with mood, coping, and immune change in response to stress. J Pers Soc Psychol 1998;74:1646 –1655. 28. Byrnes DM, Antoni MH, Goodkin K, et al. Stressful events, pessimism, natural killer cell cytotoxicity, and cytotoxic/suppressor T cells in HIV� black women at risk for cervical cancer. Psychosom Med 1998;60:714 – 722. 29. Coyne JC, Pajak TF, Harris J, et al. Emotional well-being does not predict survival in head and neck cancer patients: a Radiation Therapy Oncology Group study. Cancer 2007;110:2568 –2575. 30. Hermann BP, Trenerry MR, Colligan RC. Learned helplessness, attributional style, and deprssion in epilepsy. Epilepsia 1996;37:680 – 686. 31. Malinchoc M, Rocca WA, Colligan RC, et al. Premorbid personality characteristics of Alzheimer’s disease: an exploratory case-control study. Eur J Neurol 1997;4:227–230. 32. Ames SC, Vickers KS, Decker PA, et al. Select Minnesota Multiphasic Personality Inventory (MMPI) scales as predictors of tobacco abstinence following treatment for nicotine dependence. Psychol Health 2005;20: 331–351. 33. Brummett BH, Helms MJ, Dahlstrom WG, et al. Prediction of all-cause mortality by the Minnesota Multiphasic Personality Inven tory Optimism-Pessimism Scale scores: study of a college sample during a 40-year follow-up period. Mayo Clin Proc 2006;81:1541– 1544. 34. Grossardt BR, Bower JH, Geda YE, et al. Pessimistic, anxious, and depressive personality traits predict all-cause mortality: the Mayo Clinic cohort study of personality and aging. Psychosom Med 2009;71:491– 500.
Copyright © 2010 by the International Association for the Study of Lung Cancer
A Pessimistic Explanatory Style Is Prognostic for Poor Lung
Cancer Survival Paul Novotny, MS,* Robert C. Colligan, PhD,† Daniel W. Szydlo, BA,* Matthew M. Clark, PhD,†
Sarah Rausch, PhD,† Jason Wampfler, BS,‡ Jeff A. Sloan, PhD,* and Ping Yang, MD, PhD§
Background: Several studies have demonstrated the importance of personality constructs on health behaviors and health status. Having a pessimistic outlook has been related to negative health behaviors and higher mortality. However, the construct has not been well explored in cancer populations. Methods: Survival time of 534 adults who were diagnosed with lung cancer was examined. The patients had completed the Minnesota Multiphasic Personality Inventory approximately 18.2 years before receiving their lung cancer diagnosis. Minnesota Multiphasic Personality Inventory Optimism-Pessimism scores were divided into high (60 or more) and low scores (�60), and log-rank tests and Kaplan-Meier curves were used to determine survival differences. Multivariate Cox models were used for assessing prognostic values of pessimism along with other known predictors for lung cancer survival outcome. Bootstrapping of the survival models was used as a sensitivity analysis. Results: At the time of lung cancer diagnosis, patients were at an average age of 67 years old; 48% of them were women, 85% had non-small cell lung cancer, 15% had small cell lung cancer, 30% were stage I, 4% were stage II, 31% were stage III/limited, and 35% were stage IV/extensive. Patients who exhibited a nonpessimistic explanatory style survived approximately 6 months longer than patients classified as having a pessimistic explanatory style. Conclusion: Among lung cancer patients, those having a pessimistic explanatory style experienced a less favorable survival outcome, which may be related to cancer treatment decisions. Further research in this area is warranted. Key Words: Explanatory style, Optimism, Pessimism, Lung cancer, MMPI, Survival.
M
“
ind-body” relationships have been revered since the time of Socrates. Personality or emotional factors may have a direct impact on physiological states or mind, which are evident in the current emphasis on stress reduction, relaxation, meditation, and activities related to disease pre vention and wellness promotion. Indeed, psychosocial factors may be predictive of poor disease outcome, including cancer survivorship. Pessimism and optimism are personality con structs that have been shown to be important in the general population and in some medical populations.1,2 Having a pessimistic explanatory style means the individual attributes bad events to internal, stable, and global causes; he or she tends toward self-blame, fatalism, and catastrophic thinking.3 Recently, the role of having a pessimistic explanatory style has been explored in several cancer populations. For exam ple, Kung et al.4 found that having a pessimistic explanatory style was associated with poor quality of life in head and neck and thyroid cancer survivors; Petersen et al.5 found that a pessimistic explanatory style was predictive of poor quality of life in 268 breast cancer survivors. In contrast, optimism has also been associated with positive health outcomes in a number of studies; for instance, higher levels of optimism have been associated with lower blood pressure during daily life,6 better recovery from coronary artery bypass surgery,7 and longer survival in patients with head and neck cancer.8 However, such factors may be of little importance when the body experiences serious or fatal medical conditions, or potentially lethal diseases such as lung cancer. Using Selig man’s theory of causal attribution,3 the current retrospective study examined the relationship between a pessimistic ex planatory style and survival in a group of patients diagnosed with lung cancer.
(J Thorac Oncol. 2010;5: 326–332)
PATIENTS AND METHODS Study Design and Patients *Cancer Center Statistics, †Department of Psychiatry and Psychology, ‡Di vision of Biomedical Statistics and Informatics, and §Division of Epide miology, Mayo Clinic College of Medicine, Rochester, Minnesota. Disclosure: The authors declare no conflict of interest. Address for correspondence: Ping Yang, MD, PhD, Division of Epidemiol ogy, Mayo Clinic, 200 First Street SW, Rochester, MN 55905. E-mail: [email protected] Copyright © 2010 by the International Association for the Study of Lung Cancer ISSN: 1556-0864/10/0503-0326
326
This study is a retrospective, observational cohort design aimed at examining the relationship between ex planatory style, as measured by scores on the OptimismPessimism (PSM) scale of the Minnesota Multiphasic Personality Inventory (MMPI), and survival in patients diagnosed with lung cancer. Participants in this study were patients with primary lung cancer who were enrolled into the Epidemiology and Genetics of Lung Cancer Research Program at Mayo Clinic Rochester since 1997, in which all patients at Mayo Clinic who were diagnosed with lung
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
cancer have been offered enrollment in a prospective cohort study.9 –12 All patients enrolled provided informed consent, and the study has been approved by the Institu tional Review Board. Trained study personnel reviewed the medical records; all patients completed health-related surveys when they entered the cohort study and again at 6 and 12 months and were mailed similar surveys on an annual basis. Information on demographics, previous or concurrent illnesses, tobacco usage and exposure, tumor staging, nutritional habits, and cancer therapy were ab stracted and entered into the database. Comorbidities were combined into three variables for having any other lung disease, any other cancer, and any other disease. Having another lung disease included asthma, chronic bronchitis, emphysema, chronic obstructive pulmonary disease, pneu monia, tuberculosis, or cystic fibrosis. Any other cancer was defined as a cancer diagnosis except lung cancer. Any other disease was defined as any disease except cancer or a lung disease; this category included diabetes, heart disease, arthritis, anemia, migraines, drug addiction, and any other medical problem. For a more detailed discussion of variable definitions, see Ref. 9. The Revised Staging System of non-small cell lung cancer (NSCLC) was used.14 All cancer treatment decisions were deferred to the individual patient’s healthcare providers, and enrollment into the research program did not in any way influence clinical decision making. This study used retrospective data from 534 individuals diagnosed with lung cancer who had completed the MMPI before receiving their lung cancer diagnosis. Medical records included information on age, sex, disease stage, smoking status, and survival status. If patients completed more than one MMPI, the PSM score from the earliest MMPI completed before lung cancer diagnosis was used to assess PSM. The MMPI was com pleted at the request of the patient’s physicians as part of the patient’s medical care or as a participant in a large research study. Patients who were asked to complete an MMPI were more likely among the most complex of referrals at our tertiary care medical center, because the physician making the request believed there were psycho logic factors intertwined with the presenting symptoms.
Measures Smoking Status Patients who reported smoking fewer than 100 ciga rettes in their lifetime were classified as “never smokers.” Patients who had not smoked a cigarette in the past 30 days were classified as “former smokers,” and patients who re ported smoking any cigarette in the past 30 days were classified as “current smokers.”15
The Minnesota Multiphasic Personality Inventory The original MMPI consisted of 550 unique true/ false items about thoughts, feelings, attitudes, physical symptoms, emotional symptoms, and previous life experi ences.16 The original MMPI and its current revision, the MMPI-2,17 have been the most widely used and thoroughly researched of the self-report measures of personality func
Pessimistic Explanatory Style for Lung Cancer
tioning.18 All the analyses in the article were based on the original MMPI. The PSM scale for the MMPI was derived from Seligman’s theory of explanatory style19 and constructed from the items of the MMPI item pool.20 According to Seligman’s theory, the manner in which individuals ex plain the cause of significant life events (both good and bad events) exerts considerable influence over three as pects of their future physical and mental health, which include decreased quality of physical health, increased risk for depression, and reduced occupational or aca demic achievement.21 More specifically, Seligman’s the ory postulates that people who (1) attribute the causes of adverse events in their lives to themselves (i.e., an interval explanation, “It’s me . . .”), (2) carry the expectation that the condition will persist (i.e., a stable explanation, “. . . happened again, as usual . . .”), and (3) that it will affect other aspects of their functioning (e.g., a global explanation, “. . . and now my life will be ruined; I’ll never get to . . .”) are at risk for undermining their subse quent physical health, emotional and mental functioning, and life achievement. The PSM scale yields normalized T-scores (mean � 50, SD � 10). Lower PSM scores indicate having an optimistic explanatory style; whereas, higher scores indi cate a pessimistic explanatory style. Similar to previous research, PSM scores were divided into two groups for analysis: PSM scores representing the nonpessimistic (e.g., optimistic) patients were in the first group (PSM scores of �60). Subjects with PSM scores of 60 or higher (1 SD above the mean) were considered to have a pessimistic explanatory style.4,5 Although this cut point has been used in other studies of pessimists,1,2,4 it is not based on a clinical cutoff score. We used recursive partitioning22 to determine the appropriateness of the score as our cut point when modeling survival times; the optimal cut point for our study cohort fell at 58, but we used a normalized score of 60 as being nearly optimal for the purpose of compa rability with published literature.
Analyses PSM scores were divided into pessimistic explana tory style or nonpessimistic explanatory style, and t tests were used to determine demographic differences between patients highest and lowest in PSM scores. Log-rank tests and Kaplan-Meier curves were used to determine survival differences. Stepwise Cox proportional hazards models were used to model survival adjusting for age, gender, smoking status, cancer type, stage, and comorbidities. Variables were entered into the model one at a time. Modeling stopped when no other variables met the 0.05 significance level for entry into the model. Backward modeling (starting with a saturated model and dropping one variable at a time) was used to confirm the results of the stepwise model. Martingale residuals were explored to determine the functional form of the PSM variable, and diagnostic checks were conducted to verify the fit of the final models. Bootstrapping of the survival analysis models was carried out as a sensitivity test of our findings by
Copyright © 2010 by the International Association for the Study of Lung Cancer
327
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Novotny et al.
TABLE 1. Selected Characteristics of Patients with Lung Cancer Who Completed the MMPI Before Being Diagnosed with Lung Cancer Characteristicsa Age at DXc Mean (SD) Stage I II III/limited IV/extensive Cancer type SCLCd NSCLCe Median time from diagnosis to last follow-up
MMPIb (N � 534)
No MMPIb (N � 8786)
67.4 (10.39)
65.7 (11.14)
147 (30) 21 (4) 152 (31) 176 (36)
1999 (23) 589 (7) 2659 (31) 3310 (39)
72 (15) 424 (85) 14.8 mo
933 (11) 7624 (89) 15.5 mo
a
Unless specified, number and percentage in parentheses are presented. MMPI, Minnesota Multiphasic Personality Inventory; c DX, Diagnosis; d SCLC, small cell lung cancer; eNSCLC, non-small cell lung cancer. b
generating 10,000 random samples with replacement from our observed dataset.23
RESULTS A total of 534 subjects had completed a MMPI before receiving their lung cancer diagnosis between 1997 and 2006. Table 1 shows patients with and without a MMPI. There is no indication that patients with MMPI scores were clinically different on these characteristics than patients who did not complete a MMPI, and explan atory style is unlikely to have biased completion rates for the MMPIs among medical outpatients.24 Table 2 provides baseline demographics for the 534 patients who form the basis of this report. On average, patients had completed the MMPI 18.2 years before being diagnosed with lung cancer. Most patients had NSCLC (85%), and 34% of patients with NSCLC had early stage (I or II) lung cancer. In addition, most patients were current or former smokers (80%). None of the demographic variables were significantly different between patients with a pessimistic explanatory style and patients with a nonpessimistic explanatory style. There were 110 patients who completed more than one MMPI before their lung cancer diagnosis. Of the 110 patients with multiple MMPI scores, 12 (11%) went from being classified as pessimists to nonpessimists and 13 (12%) changed from nonpessimists to pessimists. The median change for the patients was a decrease of one-half point on their PSM scores. We used the first PSM score reported before lung cancer diagnosis to achieve a greater temporal span between the PSM score and cancer diagno sis. This minimizes the likelihood that a PSM score was affected by proximity to the personal stresses of receiving a diagnosis of lung cancer and the significant symptoms of illness preceding the cancer diagnosis related to the cancer.
328
TABLE 2. Characteristics of 534 Patients with Lung Cancer Who Completed a MMPI
Characteristicsa Age at MMPIb Mean (SD) Median Range Age at DXc Mean (SD) Median Range Years From MMPIb to DX Mean (SD) Median Range Smoking status Missing Never smoker Former smoker Current smoker Current or former smoker Gender Female Male Cancer type SCLCe NSCLCf Stage Missing I II III/limited IV/extensive Any surgery Yes No Any chemotherapy Yes No Any radiation therapy Yes No Any other cancer Yes No Any other lung disease Yes No
Nonpessimistic Pessimistic Explanatory Style Explanatory PSMd <60 Style PSMd >60 (n � 304) (n � 230)
Total (n � 534)
48.4 (12.94) 49 14–97
49.9 (12.49) 51 15–78
49.0 (12.76) 50.0 14–97
67.6 (11.10) 68.5 34–98
67.0 (9.40) 68.0 39–88
67.4 (10.39) 68 34–98
19.0 (10.08) 18.6 0.01–43.5
17.1 (9.38) 16.6 0.02–43.4
18.2 (9.82) 17.5 0.01–43.5
1 48 (16) 117 (39) 117 (39) 21 (7)
2 21 (9) 89 (39) 103 (45) 15 (7)
3 69 (13) 206 (39) 220 (41) 36 (7)
149 (49) 155 (51)
108 (47) 122 (53)
257 (48) 277 (52)
37 (13) 243 (87)
35 (16) 181 (84)
72 (15) 424 (85)
24 82 (28) 10 (4) 94 (34) 94 (34)
14 65 (30) 11 (5) 58 (27) 82 (38)
38 147 (30) 21 (4) 152 (31) 176 (36)
123 (40) 181 (60)
75 (33) 155 (67)
198 (37) 336 (63)
85 (28) 219 (72)
62 (27) 168 (73)
147 (28) 387 (72)
69 (23) 235 (77)
40 (17) 190 (83)
109 (20) 425 (80)
32 (11) 272 (89)
34 (15) 196 (85)
66 (12) 468 (88)
61 (20) 243 (80)
53 (23) 177 (77)
114 (21) 420 (79)
a
Unless specified, number and percentage in parentheses are presented. MMPI, Minnesota Multiphasic Personality Inventory; c DX, Diagnosis; PSM, Optimism-Pessimism Scale Score; eSCLC, small cell lung cancer; f NSCLC, non-small cell lung cancer. b
d
Copyright © 2010 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
TABLE 3.
Median Survival (mo) by PSM Groups 95% Confidence Interval
Overall population PSMa �60 (nonpessimistic) 60 � PSMa (pessimistic) a
Median
Lower
Upper
14.8 19.2
13.4 14.8
18.7 23.5
13.1
10.9
14.8
TABLE 4.
5-Year Log-Rank Survival p Rate 0.0105
27.8% 32.9% 21.1%
PSM, Optimism-Pessimism Scale Score.
Pessimistic Explanatory Style for Lung Cancer
Univariate Cox Models for Survival
Variable Pessimist (60�) Age at diagnosis Male Stage II Stage III/limited Stage IV/extensive Former smoker Current smoker Current or former smoker Chemotherapy Radiation Surgery Other lung treatment Any cancer Any lung disease Any other disease
95% Hazard Ratio Confidence Limits
p
Hazard Ratio
0.0108 �0.0001 0.0141 0.1500 0.2087 �0.0001 0.0452 0.0045 0.0034
1.305 1.022 1.292 0.655 1.155 4.134 0.805 1.346 1.706
1.064 1.012 1.053 0.368 0.923 3.305 0.651 1.097 1.193
1.601 1.033 1.585 1.165 1.445 5.171 0.995 1.653 2.441
0.0058 0.0253 �0.0001 0.2965
1.365 1.317 0.234 1.601
1.094 1.035 0.182 0.662
1.703 1.676 0.300 3.870
0.0675 0.0012 �0.0001
0.748 0.655 0.391
0.549 0.506 0.293
1.021 0.846 0.521
TABLE 5. Multivariate Cox Model for Survival Including Pessimism, Treatment, and Comorbidities
Variable
FIGURE 1. Kaplan-Meier curve of survival time by PSM groups. Patients with a pessimistic explanatory style have shorter survival times than other patients.
Pessimistic Explanatory Style and Survival Table 3 shows median survival times (in months) and the 5-year survival rate for all 534 patients and compares patients with low versus high scores on the PSM scale. Patients (both women and men) in the nonpessimistic category survived about 6 months longer compared with patients with a pessimistic explanatory style, as shown in the Kaplan-Meier curve in Figure 1. Five-year survival rates for the two groups were 32.9% for nonpessimists and 21.1% for pessimists. The results from univariate Cox models are shown in Table 4. Stage of disease and treatment were highly correlated, i.e., 95% of patients with stage II cancer had surgery, 30% of patients with stage III had surgery, and only 6% of patients with stage IV had surgery. Because there was strong collinearity between stage and treatment, only treatment was used in the multivariate models along with other covariates. Results from the stepwise multivar iate model using pessimists versus nonpessimists are shown in Table 5, confirming that patients with a pessi mistic explanatory style have significantly worse survival rates even after adjusting for other known prognostic variables (hazard ratio � 1.25, p � 0.03). Having comor bidities seem to be associated with significantly better survival (Table 4); however, the effects are confounded by
Pessimist (60�) Age at diagnosis Male Any surgery Any radiation therapy Any chemotherapy Current smoker Some smoking history Any other cancer Any other lung disease
p
Hazard Ratio
0.0339 �0.0001 0.02 �0.0001 0.17 0.70 0.0005 0.0005 0.35 0.66
1.25 1.04 1.28 0.23 1.22 0.95 1.54 2.02 1.18 1.07
95% Hazard Ratio Confidence Limits 1.02 1.03 1.04 0.17 0.92 0.73 1.21 1.36 0.83 0.80
1.55 1.05 1.58 0.31 1.63 1.24 1.96 3.01 1.66 1.42
age, smoking status, and treatment. In particular, patients with comorbidities were more likely to have surgery and more likely to be current smokers. Patients with other cancers were on average about 5 years older than patients without other cancers. After adjusting for these confound ing effects as presented in Table 5, the effects of comor bidities are no longer significant. A stratified analysis was used to assess the relationship between pessimism and stage. Results suggest that having a pessimistic explanatory style is prognostic for survival in patients with stage I or II cancer but is not prognostic for patients with stage III or IV cancer. The observation was supported by a multivariate Cox model using only patients with stage I or II lung cancer (Table 6). After adjusting for demographics, comorbidities, and treatment, having a pes simistic explanatory style was still significantly related to
Copyright © 2010 by the International Association for the Study of Lung Cancer
329
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Novotny et al.
TABLE 6. Multivariate Cox Model for Survival for Patients with Stage I or II Cancer
Variable Pessimist (60�) Age at diagnosis Male Any surgery Any radiation therapy Any chemotherapy Current smoker Some smoking history Any other cancer Any other lung disease
95% Hazard Ratio Confidence Limits
p
Hazard Ratio
0.02 0.001 0.17 0.01 0.10 0.64 0.03 0.10
1.91 1.06 1.47 0.39 1.95 1.21 1.87 2.52
1.10 1.02 0.85 0.19 0.87 0.53 1.06 0.84
3.29 1.10 2.53 0.81 4.36 2.76 3.31 7.57
0.33 0.25
0.72 1.42
0.38 0.79
1.38 2.56
shorter survival times (log-rank test p � 0.02, hazard ratio � 1.91). Several different sensitivity analyses provided results that were similar to the original analyses. A simulation involving 10,000 bootstrapped samples showed the robust ness of the results. Seventy-two percent of the bootstrapped samples resulted in significant p values, with nonpessimists having longer survival times than pessimists. A small minor ity of samples (0.4% or 42 out of 10,000 samples) had results that showed pessimists having longer survival times, but the survival differences were not significant in all of those samples. Another sensitivity analysis used the last MMPI score reported before diagnosis instead of the first reported MMPI. In this analysis, the log-rank p value for differences in survival increased slightly and the difference in median survival decreased from 6.0 to 5.3 months. The relationship between depression (as measured by the MMPI, Scale 2, Depression) and survival was also ex plored, but there was no indication that depression was related to survival time (data not shown).
DISCUSSION This retrospective study examined the relationship be tween survival and explanatory style using scores from the PSM scale of the MMPI. The major finding in the study was that explanatory style obtained from a large sample of pa tients diagnosed with lung cancer and measured (on average) 18 years before receiving their lung cancer diagnosis, was statistically significant and clinically relevant for survival. More specifically, patients classified as having an optimistic or a nonpessimistic explanatory style survived an average of 6 months longer compared with the patients with a pessimis tic explanatory style. Furthermore, the relationship was inde pendent of smoking status, cancer stage, treatment, comor bidities, age, and gender. In examining stage of cancer, the prolonged survival time was only sustained among patients with stage I or II lung cancer, which indicates that if a patient has advanced disease,
330
then the potentially protective aspect of a nonpessimistic explanatory style can be overwhelmed by the severity of the disease process. However, the results support the premise that if a patient is diagnosed with lung cancer, and if the patient has a pessimistic explanatory style, the patient may be less likely to have surgery and may live 6 months less compared with patients with a nonpessimistic style. This may be due to the overall risk for comorbid medical problems (i.e., as predicted by Seligman’s theory of pessimistic causal attribu tion), which in turn decreases the likelihood of receiving a surgical intervention. However, if a patient’s lung cancer was identified in the early stage, and if the patient has a nonpes simistic explanatory style, then the patient’s estimated sur vival is significantly improved, in part, because surgical resection of the tumor can be completed as a curative treat ment. This 6-month potential benefit related to an optimistic explanatory style is more impressive when one considers that the median survival time for this patient population with lung cancer is �1 year.9 Seligman’s theory of explanatory style has been shown to be predictive of health status, mortality, health behaviors, and quality of life among general medical outpatients. The examination of a possible relationship between a pessimistic explanatory style and survivorship in oncology populations is a relatively new and provocative area of investigation. Such studies have yielded mixed results. Some suggest that having a pessimistic explanatory style before receiving a cancer diagnosis might be predictive of survival time and immune function; whereas, others have not found such an association. Rausch et al.25 found that higher levels of optimism were associated with improved immune function in women newly diagnosed with early stage breast cancer. Even in healthy subjects, researchers have found optimism to be associated with higher immune parameters, including higher T-lympho cyte numbers and natural killer cell activity.26,27 In a popu lation of black women coinfected with human immunodefi ciency virus and human papillomavirus, Byrnes et al.28 found that women who were more optimistic had better immunity. Specifically, greater optimism was related to higher natural killer cell cytotoxicity and cytotoxic/suppressor cell numbers after controlling for presence/absence of human papillomavirus, behavioral/lifestyle factors, and subjective impact of negative life events. In a recent review article, Coyne et al.29 noted mixed results from studies (often flawed by small samples and numerous potential confounders) examining broad person ality factors related to disease progression and mortality among patients with cancer. In summarizing the literature, Coyne et al. decry the lack of substantial support for widely held beliefs about the impact of personality factors, using a research model of their own large-scale work in head and neck cancers. Contrary to the view of Coyne et al., we deliberately chose not to use a broad measure of emotional well being. Rather, we used a theory-derived measure characterizing an enduring psychologic and cog nitive construct known as causal attribution. A summary of our approach follows:
Copyright © 2010 by the International Association for the Study of Lung Cancer
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
Originally, Mayo Clinic researchers collaborated with Seligman and, using Seligman’s procedures for analyzing expository text, developed and published a bidirectional scale of PSM derived from the item pool of the original MMPI. Thus, Seligman’s theory of causal attribution was operation alized in the constructs of durable “pessimistic” and “opti mistic” personality traits. According to Seligman’s theory, individuals with a pessimistic explanatory style are at signif icant risk for later problems in three important areas of life functioning: (a) greater likelihood of adverse medical condi tions, (b) proneness toward mental health issues (particularly depression), and (c) reduced achievement (either occupa tional or academic). Seligman’s theory and the PSM scale have been vali dated among general medical outpatients. For example, a pessimistic explanatory style was significantly associated with increased mortality among medical outpatients who completed the MMPI approximately 30 years before follow up.1 Overall, results from several previous studies30 –34 have added considerable support to Seligman’s theory and the potentially adverse impact of having a pessimistic explana tory style. In general, we have found, as the theory predicts, patients classified as having a pessimistic explanatory style are at risk for poorer medical outcomes; whereas, being in the nonpessimist range of scores seems to be a psychologic protective factor. Seligman’s theory predicts that patients with lung cancer who were classified as having a pessimistic explanatory style would be at risk for gradually accruing adverse medical conditions over time, which may be related to the decreased likelihood of surgical treatment for patients in this explanatory category. Note: we are not implying a causal relationship between a pessimistic explanatory style and the development of lung cancer. Some limitations of the study should be considered when interpreting the results. First, the retrospective design of this study does not ensure that the measures were given at consistent times or to all possible participants. Patients who completed a MMPI might have exhibited behaviors or coping styles different from individuals who did not complete the MMPI; thereby, introducing selection bias into the study sample. Another potential source of bias relates to the mea surement of PSM at only one point in time. There is uncer tainty about whether a relationship existed between explan atory style and survival before diagnosis. A related limitation is that information was not available on cancer recurrence or comorbid medical conditions that may have occurred since the time of cancer diagnosis and could impact survival. Therefore, longitudinal studies are recommended to further investigate these possibilities. A final limitation is the gener alizability of the results; the sample was primarily white and only consisted of patients diagnosed with lung cancer, there fore, the results may not apply to more diverse populations or other types of cancer. Despite the limitations, our study adds to the grow ing literature on explanatory style (e.g., PSM) and survival in the general population, in medical populations, and specifically in patients with lung cancer. Future investiga tions may benefit from designing and testing interventions,
Pessimistic Explanatory Style for Lung Cancer
which address enhancing positive aspects of explanatory style and evaluating the potential physiological mecha nisms responsible for increased survival. For example, patients diagnosed with cancer could learn cognitive be havioral techniques to challenge negative thinking patterns and engage in effective, accurate problem solving. This may ultimately aid in enhancing current approaches to patient care, such that clinicians may improve survival not only by developing new medical treatments but also by targeting patients’ psychosocial characteristics most likely to negatively affect cancer treatment decisions and ulti mate outcomes.
ACKNOWLEDGMENTS This work was supported by U.S. National Institutes of Health grants, R01 CA 115879 and R01 CA, 84354, awarded to Ping Yang, MD, PhD. The authors thank Susan Ernst, MA, for her technical assistance with the article. REFERENCES 1. Maruta T, Colligan RC, Malinchoc M, et al. Optimists vs pessimists: survival rate among medical patients over a 30-year period. Mayo Clin Proc 2000;75:140 –143. 2. Maruta T, Colligan RC, Malinchoc M, et al. Optimism-pessimism assessed in the 1960s and self-reported health status 30 years later. Mayo Clin Proc 2002;77:748 –753. 3. Peterson C, Seligman ME. Causal explanations as a risk factor for depression: theory and evidence. Psychol Rev 1984;91:347–374. 4. Kung S, Rummans TA, Colligan RC, et al. Association of optimismpessimism with quality of life in patients with head and neck and thyroid cancers. Mayo Clin Proc 2006;81:1545–1552. 5. Petersen LR, Clark MM, Novotny P, et al. Relationship of optimismpessimism and health-related quality of life in breast cancer survivors. J Psychosoc Oncol 2008;26:15–32. 6. Raikkonen K, Matthews KA, Flory JD, et al. Effects of optimism, pessimism, and trait anxiety on ambulatory blood pressure and mood during everyday life. J Pers Soc Psychol 1999;76:104 –113. 7. Scheier MF, Matthews KA, Owens JF, et al. Optimism and rehospital ization after coronary artery bypass graft surgery. Arch Intern Med 1999;159:829 – 835. 8. Allison PJ, Guichard C, Fung K, et al. Dispositional optimism predicts survival status 1 year after diagnosis in head and neck cancer patients. J Clin Oncol 2003;21:543–548. 9. Yang P, Allen MS, Aubry MC, et al. Clinical features of 5,628 primary lung cancer patients: experience at Mayo Clinic from 1997–2003. Chest 2005;128:452– 462. 10. Sun Z, Aubry MC, Deschamps C, et al. Histologic grade is an indepen dent prognostic factor for survival in non-small cell lung cancer: an analysis of 5018 hospital- and 712 population-based cases. J Thorac Cardiovasc Surg 2006;131:1014 –1020. 11. Sugimura H, Yang P. Long-term survivorship in lung cancer. A review. Chest 2006;29:1088 –1097. 12. Jatoi A, Williams B, Nichols FC, et al. Is voluntary vitamin and mineral supplementation associated with better outcome in non-small cell lung cancer? Results from the Mayo Clinic Lung Cancer Cohort. Lung Cancer 2005;49:77– 84. 13. Visbal AL, Williams BA, Nichols FC, et al. Gender differences in non-small cell lung cancer survival: an analysis of 4,618 patients diagnosed between 1997–2002. Ann Thorac Surg 2004;78:209 –215. 14. Mountain CF. Revisions in the international system for staging lung cancer. Chest 1997;111:1710 –1717. 15. Clark MM, Croghan IT, Reading S, et al. The relationship of body image dissatisfaction to cigarette smoking in college students. Body Image 2005;2:263–270.
Copyright © 2010 by the International Association for the Study of Lung Cancer
331
Novotny et al.
Journal of Thoracic Oncology • Volume 5, Number 3, March 2010
16. Hathaway S, McKinley J. The Minnesota Multiphasic Personality In ventory. Minneapolis, Minnesota: University of Minnesota, 1943. 17. Butcher JN, Dahlstrom WG, Graham JR, et al. Manual for Adminis tration and Scoring: MMPI-2, Minnesota Multiphasic Personality Inventory-2 (Revised Edition). Minneapolis: University of Minne sota, 2001. 18. Butcher JN. A Beginner’s Guide to the MMPI-2, 2nd Ed. Washington, DC: American Psychological Association, 2005. 19. Gillham JE, Shatte AJ, Reivich KJ, et al. Optimism, pessimism, and explanatory style. In EC Chang (Ed.), Optimism and Pessimism: Impli cations for Theory, Research, and Practice. Washington, DC: American Psychological Association, 2001. 20. Colligan RC, Offord KP, Malinchoc M, et al. CAVEing the MMPI for an optimism-pessimism scale: Seligman’s attributional model and the assessment of explanatory style. J Clin Psychol 1994;50: 71–95. 21. Seligman MEP. Explanatory style: predicting depression, achievement, and health. In MD Yapko (Ed.), Brief Therapy Approaches to Treating Anxiety and Depression. New York, NY: Brunner, 1989. 22. Therneau TM, Atkinson EJ. An Introduction to Recursive Partitioning Using the Rpart Routines. Technical Report #61. Rochester: Mayo Clinic, 1997. 23. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. London: Chapman & Hall, 2003. 24. McLeod TG, Costello BA, Colligan RC, et al. Personality characteristics of healthcare satisfaction survey non-respondents. Int J Health Care Qual Assur 2009;22:145–156. 25. Rausch SM, Auerbach SA, McCain NL, et al. The relationship between psychosocial and immune variables in American women with breast cancer. In the 10th International Congress of Behavioral Medicine Meeting. Tokoyo, Japan, 2008. 26. Cohen F, Kearney KA, Zegans LS, et al. Differential immune system
332
changes with acute and persistent stress for optimists vs pessimists. Brain Behav Immun 1999;13:155–174. 27. Segerstrom SC, Taylor SE, Kemeny ME, et al. Optimism is associated with mood, coping, and immune change in response to stress. J Pers Soc Psychol 1998;74:1646 –1655. 28. Byrnes DM, Antoni MH, Goodkin K, et al. Stressful events, pessimism, natural killer cell cytotoxicity, and cytotoxic/suppressor T cells in HIV� black women at risk for cervical cancer. Psychosom Med 1998;60:714 – 722. 29. Coyne JC, Pajak TF, Harris J, et al. Emotional well-being does not predict survival in head and neck cancer patients: a Radiation Therapy Oncology Group study. Cancer 2007;110:2568 –2575. 30. Hermann BP, Trenerry MR, Colligan RC. Learned helplessness, attributional style, and deprssion in epilepsy. Epilepsia 1996;37:680 – 686. 31. Malinchoc M, Rocca WA, Colligan RC, et al. Premorbid personality characteristics of Alzheimer’s disease: an exploratory case-control study. Eur J Neurol 1997;4:227–230. 32. Ames SC, Vickers KS, Decker PA, et al. Select Minnesota Multiphasic Personality Inventory (MMPI) scales as predictors of tobacco abstinence following treatment for nicotine dependence. Psychol Health 2005;20: 331–351. 33. Brummett BH, Helms MJ, Dahlstrom WG, et al. Prediction of all-cause mortality by the Minnesota Multiphasic Personality Inven tory Optimism-Pessimism Scale scores: study of a college sample during a 40-year follow-up period. Mayo Clin Proc 2006;81:1541– 1544. 34. Grossardt BR, Bower JH, Geda YE, et al. Pessimistic, anxious, and depressive personality traits predict all-cause mortality: the Mayo Clinic cohort study of personality and aging. Psychosom Med 2009;71:491– 500.
Copyright © 2010 by the International Association for the Study of Lung Cancer